CN111814386A - Method and system for guiding hypersonic flow field into BP neural network for fine processing - Google Patents

Method and system for guiding hypersonic flow field into BP neural network for fine processing Download PDF

Info

Publication number
CN111814386A
CN111814386A CN202010509101.7A CN202010509101A CN111814386A CN 111814386 A CN111814386 A CN 111814386A CN 202010509101 A CN202010509101 A CN 202010509101A CN 111814386 A CN111814386 A CN 111814386A
Authority
CN
China
Prior art keywords
flow field
neural network
refractive index
density
error
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010509101.7A
Other languages
Chinese (zh)
Other versions
CN111814386B (en
Inventor
谢锦宇
白璐
吕强
王岩坤
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xidian University
Original Assignee
Xidian University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xidian University filed Critical Xidian University
Priority to CN202010509101.7A priority Critical patent/CN111814386B/en
Publication of CN111814386A publication Critical patent/CN111814386A/en
Application granted granted Critical
Publication of CN111814386B publication Critical patent/CN111814386B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M9/00—Aerodynamic testing; Arrangements in or on wind tunnels
    • G01M9/08—Aerodynamic models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/15—Vehicle, aircraft or watercraft design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/08—Fluids
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Computer Hardware Design (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Fluid Mechanics (AREA)
  • Molecular Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Computational Mathematics (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Automation & Control Theory (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Algebra (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention belongs to the technical field of combination of sound velocity physical computation and artificial intelligence, and discloses a method and a system for guiding a sound velocity flow field into a BP neural network for fine processing, wherein a sparse flow field data sample is refined by adopting an error reverse neural network algorithm to form refractive index flow field distribution with higher data density; and refining the six-field distribution of the acoustic velocity refractive index at the window by adopting the convergence error and the number of sampling points. The method takes sparse data as a sample to be led into a neural network for training, adopts a single hidden layer to carry out propagation of the neural network, and recalculates when the number of output layers is calculated and the weight value of connection between each neuron is changed when the error of the output layer does not meet the specified requirement; when the sampling points are more dense, the refractive index distribution near the shock wave is smoother and clearer, the fidelity is obviously improved, and the method is closer to the actual situation.

Description

Method and system for guiding hypersonic flow field into BP neural network for fine processing
Technical Field
The invention belongs to the technical field of combination of sound velocity physical computation and artificial intelligence, and particularly relates to a method and a system for guiding a sound velocity flow field into a BP neural network for fine processing.
Background
At present, the research on space target striking and hypersonic aircraft is more and more concerned, when the hypersonic aircraft flies at high speed in space, surrounding air can be compressed to form high-temperature surrounding flow field shock waves, the surrounding air is ionized to form a plasma sheath, electromagnetic waves are difficult to penetrate, so that the coupling of optics and a hypersonic flow field is enabled, the complex flow field interferes with light beam transmission, wave front distortion of the light beam is caused, and the wave front distortion is called as pneumatic optical effect. Most of the light beams transmitted in the hypersonic winding flow field adopt a ray tracing method to calculate scofflaw errors, deflection angles and optical path differences and the like. The most important step is to obtain the density and refractive index distribution of the flow field.
However, a wind tunnel method is generally adopted for simulation experiments of hypersonic aircraft flight, the experiment cost is high, the data acquisition process is complicated and difficult, and when fine flow field data are needed at local positions, the situation of data sparseness occurs. Therefore, a method for refining the experimental flow field is urgently needed, so that the result of the ray tracing method for exploring the pneumatic optical effect is more accurate.
With the rapid development of computer computing power, machine learning continuously enters the public visual field, and the artificial neural network algorithm is applied to various industries. An Error Back Propagation (BP) neural network is the most widely used neural network, the output result of which is carried out in a forward Propagation manner, and the Error is carried out in a backward Propagation manner.
The existing method for reconstructing the hypersonic velocity flow field generally adopts an interpolation method or a derivation method of the interpolation method, and generally has the defects of low precision and low fidelity, or low efficiency caused by the fact that the precision meets the requirement but the calculation time is too long. And the BP algorithm is to keep the weight value to reconstruct the flow field, and the reconstruction process is rapid and has high accuracy.
Through the above analysis, the problems and defects of the prior art are as follows: (1) the existing hypersonic velocity flow field experiment has high cost and complex data sampling process, and the requirement is difficult to meet when fine flow field data are needed at the position of a flight window;
(2) for optical transmission, a fine refractive index field is often a determining factor for determining a result of calculating an aerodynamic optical effect, and if a flow field is not very fine, an accurate result is difficult to obtain;
(3) the common interpolation method is not high in accuracy for reconstructing the flow field, and if the required accuracy is achieved, a large amount of calculation time is consumed, and the efficiency is too low.
The difficulty in solving the above problems and defects is:
and butting the hypersonic flow field data with a BP algorithm to construct connection weights of the neural networks belonging to different flow fields.
The significance of solving the problems and the defects is as follows:
the method has the advantages of improving the fineness of the flow field, greatly improving the precision of a ray tracing calculation result and having great reference value in the aspect of laser guidance precision.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a method and a system for guiding a hypersonic flow field into a BP neural network for fine processing. And guiding the hypersonic flow field in sparse distribution into a BP neural network for fine processing to obtain flow field distribution with higher precision.
The method for guiding the sound velocity flow field into the BP neural network to perform refinement processing is realized by adopting an error reverse neural network algorithm, refining sparse flow field data samples to form refractive index flow field distribution with higher data density, and refining the refractive index field distribution at a window by adopting different convergence errors and sampling points.
The method specifically comprises the following steps:
(1) simulating the distribution of the density of a gas flow field of the high-speed aircraft flying at high altitude by using simulation software;
(2) the density values are converted into corresponding refractive index values. The conversion formula for converting the density into the refractive index is G-D formula, and the G-D formula is n ═ 1+ KGD·ρ；
(3) Realizing a BP neural network algorithm by using a programming technology;
(4) inputting the horizontal and vertical coordinate data obtained by simulation and the refractive index of the corresponding coordinate into a neural network algorithm;
(5) and extracting the output horizontal and vertical coordinates and the refractive indexes corresponding to the coordinates, and reconstructing the flow field to achieve the effect of refining the flow field.
Further, the step (1) further comprises:
establishing a blunt-like model, flying at high altitude under the condition of hypersonic speed, and simulating the density characteristic of the blunt-like model external flow field;
the step (2) further comprises: extracting the coordinates of the flow field obtained by simulation and the corresponding density, and converting the density value into the corresponding refractive index value by using a G-D formula; the general G-D formula is: n is 1+ KGDρ, wherein
Figure BDA0002527747380000031
Where λ is in μm.
The step (3) further comprises:
programming a BP neural network algorithm, wherein the algorithm has the main idea that training samples are input, output values are calculated by utilizing weights among neurons and an activation function, errors exist between the output values and sample values, the errors are propagated reversely, the weights are adjusted to enable the output and the samples to be close, training is finished when the mean square error is smaller than a specified error, and the weights are reserved.
The step (4) further comprises:
(4a) the flow field is obtainedTo the horizontal and vertical coordinates after normalization as sample input layers
Figure BDA0002527747380000032
Inputting a BP neural network;
(4b) refractive index of
Figure BDA0002527747380000034
Performing normalization to obtain
Figure BDA0002527747380000033
That is, the sample expectation will be associated with the output layer
Figure BDA0002527747380000035
Comparing to obtain the error between them;
Figure BDA0002527747380000036
and
Figure BDA0002527747380000037
respectively normalized values of x 'and y' in abscissa and forming an input layer of the neural network;
Figure BDA0002527747380000038
forming an output layer for outputting the refractive index; the middle part is a hidden layer, the hidden layer arranged in the example is 1 layer, the hidden layer has 30 neurons in total, and k is 30;
(4c) when calculating an output refractive index
Figure BDA0002527747380000039
It is compared with the expected value
Figure BDA00025277473800000310
If the error is larger than the specified error, the error is propagated reversely, and the weight between the neurons is readjusted;
(4d) after the weight value is adjusted, recalculating until the error between the output value and the expected value is smaller than the designated error, stopping calculating, storing the weight value connected between the neurons, and finishing training;
(4e) and entering a prediction stage after training, inputting a large number of coordinate parameters with extremely small intervals again, and outputting a refractive index value with the error corresponding to the coordinates smaller than the specified error through training the stored neural network.
The step (5) further comprises:
and performing inverse normalization on the output values and forming a refined refractive index flow field corresponding to the horizontal and vertical coordinates.
Another objective of the present invention is to provide a system for guiding a sound velocity flow field into a BP neural network for refinement, which includes:
the gas flow field density distribution simulation module simulates the distribution of the gas flow field density of the high-speed aircraft flying at high altitude by using simulation software;
the conversion formula acquisition module is used for converting the density value into a corresponding refractive index value; converting the density into a refractive index into a G-D formula;
the BP neural network algorithm editing module is used for realizing a BP neural network algorithm by using a programming technology;
the flow field refinement module is used for inputting the horizontal and vertical coordinate data obtained by simulation and the refractive index of the corresponding coordinate into a neural network algorithm; and extracting the output horizontal and vertical coordinates and the refractive indexes corresponding to the coordinates, reconstructing a flow field and realizing the refinement of the flow field.
Another object of the present invention is to provide a program storage medium for receiving user input, wherein a stored computer program causes an electronic device to execute a method for guiding the sound velocity flow field into a BP neural network for refinement, the method comprising the following steps:
step 1, simulating the distribution of the density of a gas flow field of a high-speed aircraft flying at high altitude by using simulation software;
step 2, converting the density value into a corresponding refractive index value; converting the density into a refractive index into a G-D formula;
step 3, implementing a BP neural network algorithm by using a programming technology;
step 4, inputting the horizontal and vertical coordinate data obtained by simulation and the refractive index of the corresponding coordinate into a neural network algorithm;
and 5, extracting the output horizontal and vertical coordinates and the refractive indexes corresponding to the coordinates, reconstructing a flow field and realizing the refinement of the flow field.
Another object of the present invention is to provide a hypersonic aerocraft, which carries a system for performing refinement on an acoustic flow field introduction BP neural network, where the system for performing refinement on the acoustic flow field introduction BP neural network includes:
the gas flow field density distribution simulation module simulates the distribution of the gas flow field density of the high-speed aircraft flying at high altitude by using simulation software;
the conversion formula acquisition module is used for converting the density value into a corresponding refractive index value; converting the density into a refractive index into a G-D formula;
the BP neural network algorithm editing module is used for realizing a BP neural network algorithm by using a programming technology;
the flow field refinement module is used for inputting the horizontal and vertical coordinate data obtained by simulation and the refractive index of the corresponding coordinate into a neural network algorithm; and extracting the output horizontal and vertical coordinates and the refractive indexes corresponding to the coordinates, reconstructing a flow field and realizing the refinement of the flow field.
By combining all the technical schemes, the invention has the advantages and positive effects that:
(1) the calculation precision is high, the weight is reset by adopting an error back propagation mode, and when the error is higher than a specified error, back propagation is carried out to ensure the precision;
(2) the calculation efficiency is high, only weight and error and neuron are used as main parts of the neural network in the calculation of the whole neural network, the intermediate variable is less, an input-output system is adopted instead of a circulating input-integration output system, and the calculation efficiency is ensured;
(3) the sparse flow field is distributed densely, and the originally obtained sparse data flow field is converted into a data dense flow field, so that great help is brought to subsequent solving of physical problems.
The method has an important effect on processing the hypersonic aircraft winding flow field, adopts a BP neural network algorithm to carry out fine processing on the sparse flow field obtained by experiments or simulation to obtain high-fidelity flow field distribution, and plays a foundation role in processing the hypersonic follow-up problems.
Moreover, sparse data is used as a sample and introduced into the neural network for training, a single hidden layer is adopted for propagation of the neural network, and when the number of output layers is calculated and the error of the output layers does not meet the specified requirement, the weight value connected between each neuron is changed, and the calculation is carried out again. When the normalized mean square error of the sample and the actual output is less than 10-5And meanwhile, after the mean square error is subjected to inverse normalization processing, the corresponding actual mean square error meets the requirements, a large number of coordinates are input, and refractive index distribution data corresponding to the large number of coordinates are output, so that the distribution of the refractive index field is refined, and high-fidelity refractive index field distribution is obtained.
The key point of the invention is to combine the physical result with the artificial intelligence algorithm, and the final refined result also changes for different BP algorithms. While the refractive index field is further analyzed for geometrical optical transmission characteristics, K in the G-D formula of the present inventionGDWith respect to wavelength λ, it is meant that the beam air density for each wavelength will have a different index of refraction for it, which needs to be considered before calculation. If the method is not adopted, the calculation precision is difficult to guarantee, the obtained flow field refinement degree is also difficult to guarantee, and meanwhile, the cost for achieving precision is greatly improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments of the present application will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained from the drawings without creative efforts.
Fig. 1 is a flowchart of a method for guiding a sound velocity flow field into a BP neural network to perform refinement processing according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of a method for guiding a sound velocity flow field into a BP neural network to perform refinement processing according to an embodiment of the present invention.
Fig. 3 is a flow field before refinement (110 × 110 unit flow field) provided by an embodiment of the present invention.
Fig. 4 is a refined flow field (1000 × 1000 unit flow field) diagram provided by the embodiment of the present invention.
FIG. 5 is a theoretical plot of a gradient layered refractive index field provided by an embodiment of the present invention.
FIG. 6 is a field reconstruction pattern for providing a gradient layered refractive index according to an embodiment of the present invention.
FIG. 7 is a graph of the refractive index field distribution results for an embodiment of the present invention providing high fidelity.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Aiming at the problems in the prior art, the invention provides a method and a system for guiding a sound velocity flow field into a BP neural network for fine processing, and the invention is described in detail below with reference to the accompanying drawings.
As shown in fig. 1, the method for guiding the sound velocity flow field into the BP neural network to perform refinement processing includes:
s101, simulating the distribution of the density of the gas flow field of the high-speed aircraft flying at high altitude by using simulation software.
And S102, converting the density value into a corresponding refractive index value. The conversion formula for converting the density into the refractive index is G-D formula, and the G-D formula is n ═ 1+ KGD·ρ。
S103, implementing the BP neural network algorithm by using a programming technology.
And S104, inputting the horizontal and vertical coordinate data obtained by simulation and the refractive index of the corresponding coordinate into a neural network algorithm.
And S105, extracting the output horizontal and vertical coordinates and the refractive indexes corresponding to the coordinates, and reconstructing the flow field to achieve the effect of refining the flow field.
In step S101, one piece of software is selected from a plurality of pieces of simulation software to perform fluid simulation, and a model similar to a blunt tip is first established to allow the model to fly at high altitude and in a hypersonic flight condition, so as to simulate the density characteristics of an external flow field.
In step S102, the coordinates of the flow field obtained by simulation and the corresponding density are extracted, and the density value is converted into a corresponding refractive index value by using a G-D formula. The general G-D formula is: n is 1+ KGDρ, wherein
Figure BDA0002527747380000071
Where λ is in μm.
In step S103, the BP neural network algorithm is programmed, and the main idea of the algorithm is to input training samples, calculate output values by using weights between neurons and an activation function, where there is an error between an output value and a sample value, perform back propagation on the error, adjust the weights so that the output is as close as possible to the expected value of the sample, and terminate training when the mean square error is smaller than a specified error, and retain the weights.
In step S104, the horizontal and vertical coordinate data obtained by simulation and the refractive index of the corresponding coordinate are input to a neural network algorithm. The method comprises the following steps:
(4a) normalizing the horizontal and vertical coordinates obtained from the flow field and using the normalized horizontal and vertical coordinates as a sample input layer
Figure BDA0002527747380000081
The signal is input into a BP neural network,
(4b) refractive index of
Figure BDA0002527747380000082
Performing normalization to obtain
Figure BDA0002527747380000083
That is, the sample expectation will be associated with the output layer
Figure BDA0002527747380000084
Comparing to obtain itThe error between them.
Figure BDA0002527747380000085
And
Figure BDA0002527747380000086
normalized values of the abscissa x 'and the ordinate y', respectively, constitute the input layer of the neural network.
Figure BDA0002527747380000087
To output the refractive index, an output layer is constructed. The middle part is a hidden layer, the hidden layer provided in this example is 1 layer, and the hidden layer has 30 neurons in total, i.e., k is 30.
(4c) When calculating an output refractive index
Figure BDA00025277473800000810
It is compared with the expected value
Figure BDA0002527747380000089
If the error is larger than the specified error, the error will be propagated reversely, and the weight between the neurons is adjusted again.
(4d) And after the weight value is adjusted, recalculating until the error between the output value and the expected value is smaller than the designated error, stopping calculating, storing the weight value connected between the neurons, and finishing training.
(4e) And entering a prediction stage after training, inputting a large number of coordinate parameters with extremely small intervals again, and outputting a refractive index value with the error corresponding to the coordinates smaller than the specified error through training the stored neural network.
In step S105, the output values are inverse normalized and form a regular and more refined refractive index flow field corresponding to the horizontal and vertical coordinates.
The key point of the present invention is that step S104 to step S105 combine the physical result with the artificial intelligence algorithm, and the final refinement result will also vary for different BP algorithms. While the refractive index field is further analyzed for geometrical optical transfer characteristics, step S102, G-DIn the formula KGDWith respect to wavelength λ, it is meant that the beam air density for each wavelength will have a different index of refraction for it, which needs to be considered before calculation.
The invention takes sparse data as a sample to be led into the neural network for training, adopts a single hidden layer to carry out propagation of the neural network, and recalculates when the number of output layers is calculated and the error of the output layers does not reach the specified requirement, and the weight value of the connection between each neuron is changed. When the normalized mean square error of the sample and the actual output is less than 10-5And meanwhile, after the mean square error is subjected to inverse normalization processing, the corresponding actual mean square error meets the requirements, a large number of coordinates are input, and refractive index distribution data corresponding to the large number of coordinates are output, so that the distribution of the refractive index field is refined, and high-fidelity refractive index field distribution is obtained.
The method for searching for a transcription factor binding site provided by the present invention can be implemented by other steps, and the method for searching for a transcription factor binding site provided by the present invention of fig. 1 is only one specific example.
Fig. 2 is a schematic diagram of a method for guiding a sound velocity flow field into a BP neural network to perform refinement processing according to an embodiment of the present invention.
The invention also provides a system for guiding the sound velocity flow field into the BP neural network for fine processing, which comprises the following steps:
and the gas flow field density distribution simulation module simulates the distribution of the gas flow field density of the high-speed aircraft flying at high altitude by using simulation software.
The conversion formula acquisition module is used for converting the density value into a corresponding refractive index value; the conversion formula for converting the density into the refractive index is a G-D formula.
And the BP neural network algorithm editing module is used for realizing the BP neural network algorithm by using a programming technology.
The flow field refinement module is used for inputting the horizontal and vertical coordinate data obtained by simulation and the refractive index of the corresponding coordinate into a neural network algorithm; and extracting the output horizontal and vertical coordinates and the refractive indexes corresponding to the coordinates, reconstructing a flow field and realizing the refinement of the flow field.
The effect of the method is illustrated by a practical example as follows:
1) simulation software and basic parameters used by the examples
The used simulation software is Fluent software under Ansys, and ICEM modeling and mesh subdivision are adopted.
The basic parameters of the model flight are:
flight height: 20km, flying speed: ma — 3, ambient air pressure: 5.529X 103Pa, ambient air temperature: 216.65K.
Flow field precision and refinement precision:
110 x 110 groups of sample data are refined to 1000 x 1000 within a range of 26.5cm x 110cm at the window.
2) Data results
The original precision near the window is 110 multiplied by 110 units, as shown in fig. 3, after sampling points are more dense, the refractive index distribution near the shock wave is smoother and clearer, the fidelity is obviously improved, when the mean square error is reduced, the refractive index field is greatly refined, the hierarchical structure near the shock wave is more obvious, the refractive index near the window is also obviously divided, the actual situation is more close, the data is more fidelity, as shown in fig. 4, the distribution situation of the refractive index is reproduced to a great extent, and the data is denser than original sample data, so that effective data guarantee is provided for the accuracy of the follow-up pneumatic optical effect study.
The present invention will be further described with reference to effects.
The invention provides a method for performing fine processing on a flow field by using an error feedback neural network algorithm. And introducing sparse data serving as a sample into a neural network for training, adopting a single hidden layer for propagation of the neural network, and recalculating when the number of output layers is calculated and the weight connected between each neuron is changed when the error of the output layer does not meet the specified requirement. When the normalized mean square error of the sample and the actual output is less than 10-5Then, after the mean square error is processed by inverse normalization, the corresponding actual mean square error meets the requirement, a large number of coordinates are input at the same time, and a large number of coordinates are output to correspond to each otherThe refractive index distribution data refines the distribution of the refractive index field and obtains the high-fidelity refractive index field distribution. The partial data results obtained are shown in FIG. 7.
Firstly, a refractive index layered refractive index field is constructed, the total number of the refractive index layered refractive index field is 10, the size of a flow field area is 20cm multiplied by 20cm, the refractive index of each layer is increased by 0.05 and is increased from 1.05 to 1.5, as shown in figure 5, 10000 sampling points are extracted from a theoretical field, the method used by the invention is introduced into a BP neural network for carrying out refinement treatment and reconstructing the refractive index field, the obtained result is shown in figure 6, and the comparison of the two figures shows that the refractive index field constructed by using a BP neural network algorithm has higher refinement degree than the original true refractive index field, the hierarchy is clearer, and the respective precision of each layer is ensured. The correctness and the accuracy of the method are verified.
In the description of the present invention, "a plurality" means two or more unless otherwise specified; the terms "upper", "lower", "left", "right", "inner", "outer", "front", "rear", "head", "tail", and the like, indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, are only for convenience in describing and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed in a particular orientation, and be operated, and thus, should not be construed as limiting the invention. Furthermore, the terms "first," "second," "third," and the like are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
It should be noted that the embodiments of the present invention can be realized by hardware, software, or a combination of software and hardware. The hardware portion may be implemented using dedicated logic; the software portions may be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or specially designed hardware. Those skilled in the art will appreciate that the apparatus and methods described above may be implemented using computer executable instructions and/or embodied in processor control code, such code being provided on a carrier medium such as a disk, CD-or DVD-ROM, programmable memory such as read only memory (firmware), or a data carrier such as an optical or electronic signal carrier, for example. The apparatus and its modules of the present invention may be implemented by hardware circuits such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of hardware circuits and software, e.g., firmware.
The above description is only for the purpose of illustrating the present invention and the appended claims are not to be construed as limiting the scope of the invention, which is intended to cover all modifications, equivalents and improvements that are within the spirit and scope of the invention as defined by the appended claims.

Claims (10)

1. A method for guiding a sound velocity flow field into a BP neural network for fine processing is characterized in that the method for guiding the sound velocity flow field into the BP neural network for fine processing adopts an error inverse neural network algorithm to fine sparse sound velocity flow field data samples to form sound velocity refractive index flow field distribution with higher data density degree;
and refining the six-field distribution of the acoustic velocity refractive index at the window by adopting the convergence error and the number of sampling points.
2. The method for guiding the sound velocity flow field into the BP neural network for refinement processing as claimed in claim 1, wherein the method for constructing the sound velocity refractive index flow field distribution with higher data density further comprises:
guiding sparse sound velocity data serving as a sample into a neural network for training, adopting a single hidden layer for propagation of the neural network, and recalculating when the number of output layers is calculated and the weight connected between each neuron is changed when the error of the output layer does not meet the specified requirement; when the normalized mean square error of the sample and the actual output is less than 10-5Then, after the mean square error is processed by inverse normalization, the corresponding actual mean square errorAnd a large number of coordinates are input simultaneously, and refractive index distribution data corresponding to the large number of coordinates are output.
3. The method for performing refinement processing by introducing the sound velocity flow field into the BP neural network according to claim 1, wherein the method for performing refinement processing by introducing the sound velocity flow field into the BP neural network further comprises:
simulating the distribution of the density of a gas flow field of a high-speed aircraft flying at high altitude by using simulation software;
step two, converting the density value into a corresponding refractive index value; converting the density into a refractive index into a G-D formula;
thirdly, implementing a BP neural network algorithm by using a programming technology;
inputting the horizontal and vertical coordinate data obtained by simulation and the refractive index of the corresponding coordinate into a neural network algorithm;
and fifthly, extracting the output horizontal and vertical coordinates and the refractive indexes corresponding to the coordinates, reconstructing the flow field and realizing the refinement of the flow field.
4. The method for guiding the sonic flow field into the BP neural network for refinement as set forth in claim 3, wherein the step one further comprises:
establishing a blunt-like model, flying at high altitude under the condition of hypersonic speed, and simulating the density characteristic of the blunt-like model external flow field;
the second step further comprises: extracting the coordinates of the flow field obtained by simulation and the corresponding density, and converting the density value into the corresponding refractive index value by using a G-D formula; the general G-D formula is: n is 1+ KGDρ, wherein
Figure FDA0002527747370000021
Where λ is in μm.
5. The method for performing refinement processing by introducing the sonic flow field into the BP neural network according to claim 3, wherein the third step further comprises:
programming a BP neural network algorithm, wherein the algorithm has the main idea that training samples are input, output values are calculated by utilizing weights among neurons and an activation function, errors exist between the output values and sample values, the errors are propagated reversely, the weights are adjusted to enable the output and the samples to be close, training is finished when the mean square error is smaller than a specified error, and the weights are reserved.
6. The method for performing refinement processing by introducing the sonic flow field into the BP neural network according to claim 3, wherein the step four further comprises:
(4a) normalizing the horizontal and vertical coordinates obtained from the flow field and using the normalized horizontal and vertical coordinates as a sample input layer
Figure FDA0002527747370000022
Inputting a BP neural network;
(4b) refractive index of
Figure FDA0002527747370000023
Performing normalization to obtain
Figure FDA0002527747370000024
That is, the sample expectation will be associated with the output layer
Figure FDA0002527747370000025
Comparing to obtain the error between them;
Figure FDA0002527747370000026
and
Figure FDA0002527747370000027
respectively normalized values of x 'and y' in abscissa and forming an input layer of the neural network;
Figure FDA0002527747370000028
is refracted for outputRate, constituting an output layer; the middle part is a hidden layer, the hidden layer arranged in the example is 1 layer, the hidden layer has 30 neurons in total, and k is 30;
(4c) when calculating an output refractive index
Figure FDA0002527747370000029
It is compared with the expected value
Figure FDA00025277473700000210
If the error is larger than the specified error, the error is propagated reversely, and the weight between the neurons is readjusted;
(4d) after the weight value is adjusted, recalculating until the error between the output value and the expected value is smaller than the designated error, stopping calculating, storing the weight value connected between the neurons, and finishing training;
(4e) and entering a prediction stage after training, inputting a large number of coordinate parameters with extremely small intervals again, and outputting a refractive index value with the error corresponding to the coordinates smaller than the specified error through training the stored neural network.
7. The method for performing refinement processing by introducing the sonic flow field into the BP neural network according to claim 3, wherein the fifth step further comprises:
and performing inverse normalization on the output values and forming a refined refractive index flow field corresponding to the horizontal and vertical coordinates.
8. The system for performing refinement processing on the sound velocity flow field introduced into the BP neural network according to any one of claims 1 to 7, in which the system for performing refinement processing on the sound velocity flow field introduced into the BP neural network comprises:
the gas flow field density distribution simulation module simulates the distribution of the gas flow field density of the high-speed aircraft flying at high altitude by using simulation software;
the conversion formula acquisition module is used for converting the density value into a corresponding refractive index value; converting the density into a refractive index into a G-D formula;
the BP neural network algorithm editing module is used for realizing a BP neural network algorithm by using a programming technology;
the flow field refinement module is used for inputting the horizontal and vertical coordinate data obtained by simulation and the refractive index of the corresponding coordinate into a neural network algorithm; and extracting the output horizontal and vertical coordinates and the refractive indexes corresponding to the coordinates, reconstructing a flow field and realizing the refinement of the flow field.
9. A program storage medium for receiving user input, wherein a stored computer program causes an electronic device to execute the method for guiding the sound velocity flow field into the BP neural network for refinement processing, the method comprising the following steps:
step 1, simulating the distribution of the density of a gas flow field of a high-speed aircraft flying at high altitude by using simulation software;
step 2, converting the density value into a corresponding refractive index value; converting the density into a refractive index into a G-D formula;
step 3, implementing a BP neural network algorithm by using a programming technology;
step 4, inputting the horizontal and vertical coordinate data obtained by simulation and the refractive index of the corresponding coordinate into a neural network algorithm;
and 5, extracting the output horizontal and vertical coordinates and the refractive indexes corresponding to the coordinates, reconstructing a flow field and realizing the refinement of the flow field.
10. A method for performing the refining treatment of the sound velocity flow field introduced BP neural network according to any one of claims 1 to 7 is carried out on a hypersonic aerocraft, the hypersonic aerocraft is provided with a sound velocity flow field introduced BP neural network to perform the refining treatment system, and the sound velocity flow field introduced BP neural network to perform the refining treatment system comprises:
the gas flow field density distribution simulation module simulates the distribution of the gas flow field density of the high-speed aircraft flying at high altitude by using simulation software;
the conversion formula acquisition module is used for converting the density value into a corresponding refractive index value; converting the density into a refractive index into a G-D formula;
the BP neural network algorithm editing module is used for realizing a BP neural network algorithm by using a programming technology;
the flow field refinement module is used for inputting the horizontal and vertical coordinate data obtained by simulation and the refractive index of the corresponding coordinate into a neural network algorithm; and extracting the output horizontal and vertical coordinates and the refractive indexes corresponding to the coordinates, reconstructing a flow field and realizing the refinement of the flow field.
CN202010509101.7A 2020-06-07 2020-06-07 Method and system for conducting fine treatment through BP neural network by hypersonic flow field Active CN111814386B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010509101.7A CN111814386B (en) 2020-06-07 2020-06-07 Method and system for conducting fine treatment through BP neural network by hypersonic flow field

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010509101.7A CN111814386B (en) 2020-06-07 2020-06-07 Method and system for conducting fine treatment through BP neural network by hypersonic flow field

Publications (2)

Publication Number Publication Date
CN111814386A true CN111814386A (en) 2020-10-23
CN111814386B CN111814386B (en) 2024-02-02

Family

ID=72844695

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010509101.7A Active CN111814386B (en) 2020-06-07 2020-06-07 Method and system for conducting fine treatment through BP neural network by hypersonic flow field

Country Status (1)

Country Link
CN (1) CN111814386B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112488125A (en) * 2020-11-28 2021-03-12 重庆邮电大学 Reconstruction method and system based on high-speed visual diagnosis and BP neural network

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104239659A (en) * 2013-06-07 2014-12-24 中国石油天然气集团公司 Carbon steel corrosion rate prediction method of back propagation (BP) neural network
WO2015172560A1 (en) * 2014-05-16 2015-11-19 华南理工大学 Central air conditioner cooling load prediction method based on bp neural network

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104239659A (en) * 2013-06-07 2014-12-24 中国石油天然气集团公司 Carbon steel corrosion rate prediction method of back propagation (BP) neural network
WO2015172560A1 (en) * 2014-05-16 2015-11-19 华南理工大学 Central air conditioner cooling load prediction method based on bp neural network

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
潘文婵;刘尚东;: "BP神经网络的优化研究与应用", 计算机技术与发展, no. 05 *
黄江涛;高正红;白俊强;赵轲;李静;许放;: "应用Delaunay图映射与FFD技术的层流翼型气动优化设计", 航空学报, no. 10 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112488125A (en) * 2020-11-28 2021-03-12 重庆邮电大学 Reconstruction method and system based on high-speed visual diagnosis and BP neural network
CN112488125B (en) * 2020-11-28 2021-12-14 重庆邮电大学 Reconstruction method and system based on high-speed visual diagnosis and BP neural network

Also Published As

Publication number Publication date
CN111814386B (en) 2024-02-02

Similar Documents

Publication Publication Date Title
US20180018555A1 (en) System and method for building artificial neural network architectures
He et al. Improved particle swarm optimization algorithms for aerodynamic shape optimization of high-speed train
CN112686056B (en) Emotion classification method
Chen et al. FlowGAN: A conditional generative adversarial network for flow prediction in various conditions
CN114638048A (en) Three-dimensional spray pipe flow field rapid prediction and sensitivity parameter analysis method and device
CN112597702A (en) Pneumatic modeling generation type confrontation network model training method based on radial basis function
He et al. Aerodynamic optimisation of a high-speed train head shape using an advanced hybrid surrogate-based nonlinear model representation method
CN112733473A (en) Automatic and intelligent ship free rolling attenuation numerical simulation method based on CFD
Duan et al. High-order overset flux reconstruction method for dynamic moving grids
CN111814386B (en) Method and system for conducting fine treatment through BP neural network by hypersonic flow field
Jeon et al. Target capacity filter pruning method for optimized inference time based on YOLOv5 in embedded systems
CN114818477A (en) Aircraft aerodynamic characteristic prediction method, device and medium
Tekaslan et al. Surrogate unsteady aerodynamic modeling with autoencoders and LSTM networks
CN117910128A (en) Airfoil design method and device based on depth operator network
Mallik et al. Deep neural network for learning wave scattering and interference of underwater acoustics
He et al. Design and Implementation of Embedded Real‐Time English Speech Recognition System Based on Big Data Analysis
CN115859780A (en) Power equipment vibration simulation method and system based on physical constraint neural network
Zuo et al. Fast simulation of airfoil flow field via deep neural network
CN115544882A (en) Tokamak discharge modeling system based on bidirectional long-time and short-time memory neural network
Zhixin et al. Research of improved back-propagation neural network algorithm1
Li et al. Wind turbine wake prediction modelling based on transformer-mixed conditional generative adversarial network
Duan et al. A high-order flux reconstruction method for 3D mixed overset meshes
Li et al. Investigation of Compressor Cascade Flow Using Physics-Informed Neural Networks with Adaptive Learning Strategy
Hu et al. 3DACRNN Model Based on Residual Network for Speech Emotion Classification.
Zhang et al. A flexible and efficient optimization design framework for the shape of blend-wing-body underwater glider

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant